CN109167642B - Low-overhead real-time spectrogram construction method based on historical decision data learning - Google Patents

Low-overhead real-time spectrogram construction method based on historical decision data learning Download PDF

Info

Publication number
CN109167642B
CN109167642B CN201810875938.6A CN201810875938A CN109167642B CN 109167642 B CN109167642 B CN 109167642B CN 201810875938 A CN201810875938 A CN 201810875938A CN 109167642 B CN109167642 B CN 109167642B
Authority
CN
China
Prior art keywords
spectrum
boundary
decision data
historical
spectrogram
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810875938.6A
Other languages
Chinese (zh)
Other versions
CN109167642A (en
Inventor
曾歆
高煜
黄新林
唐小伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN201810875938.6A priority Critical patent/CN109167642B/en
Publication of CN109167642A publication Critical patent/CN109167642A/en
Application granted granted Critical
Publication of CN109167642B publication Critical patent/CN109167642B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/382—Monitoring; Testing of propagation channels for resource allocation, admission control or handover

Abstract

The invention relates to a low-overhead real-time spectrogram construction method based on historical decision data learning, which comprises the following steps of: step S1: loading collected target spectrum availability historical judgment data obtained by all sensing nodes in the large-scale cognitive radio network through a single-point spectrum sensing algorithm; step S2: analyzing the loaded historical decision data by using a least square support vector machine to determine a spectrum state boundary so as to construct a spectrogram; step S3: and determining a support vector in the historical decision data through a boundary user search algorithm to optimize the spectrogram. Compared with the prior art, the invention screens out the secondary users far away from the boundary of the spectrum state through the boundary user search algorithm, and the secondary users do not need to participate in the detection of the target frequency band, thereby reducing the energy cost and the time cost for constructing the spectrogram.

Description

Low-overhead real-time spectrogram construction method based on historical decision data learning
Technical Field
The invention relates to a spectrogram construction method, in particular to a low-overhead real-time spectrogram construction method based on historical decision data learning.
Background
The problem of interruption of public network communication caused by radio interference exists in a large amount due to frequent interference in the wireless communication environment of today. Cognitive radio technology (CR) can effectively solve these problems. Cognitive radio is defined as an intelligent wireless communication technology based on Software Defined Radio (SDR). It is capable of continuously sensing the surrounding radio environment, analyzing, making decisions, dynamically selecting idle frequencies for communication, and then adaptively modifying the communication parameters within the device using Radio Knowledge Representation Language (RKRL).
The cognitive radio network consists of two types of users, namely a Primary User (PU) with authorized frequency band use authority and a Secondary User (SU) with cognitive function and capable of opportunistically using the authorized frequency band. Cognitive capabilities refer to the ability of a cognitive radio user to capture or detect useful information in the wireless environment in which the user is operating. And acquiring and analyzing the sensing data in a wireless environment, and finally obtaining a target frequency spectrum availability judgment result, namely frequency spectrum sensing. The spectrum sensing can be divided into single-point spectrum sensing and multi-point cooperative spectrum sensing, and is respectively completed by one secondary user and a plurality of secondary users. According to a traditional spectrum sensing multi-selection cooperative spectrum sensing strategy, secondary users are clustered according to spectrum states of the secondary users, and therefore usability judgment can be made on sensing information of nodes in a cluster in a unified mode. In a large-scale cognitive radio network faced by the invention, secondary user spectrum states cannot be clustered accurately due to the terrain, the positions of primary users and the like, so that a single-point spectrum sensing strategy is used on each sensing node, namely each node independently makes an availability judgment on a target frequency band according to information sensed by the node.
The current spectrum sensing is not sufficient in utilization of historical data, a large research space is provided, and introduction of a machine learning algorithm into the field of cognitive radio is an effective way for improving the performance of a cognitive radio network. In a large-scale cognitive radio network, as historical data has the characteristics of large data volume, more redundant information and the like, the time complexity for establishing a spectrogram by using the prior art is high.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a low-overhead real-time spectrogram construction method based on historical decision data learning.
The purpose of the invention can be realized by the following technical scheme:
a low-overhead real-time spectrogram construction method based on historical decision data learning comprises the following steps:
step S1: loading collected target spectrum availability historical judgment data obtained by all sensing nodes in the large-scale cognitive radio network through a single-point spectrum sensing algorithm;
step S2: analyzing the loaded historical decision data by using a least square support vector machine to determine a spectrum state boundary so as to construct a spectrogram;
step S3: and determining a support vector in the historical decision data through a boundary user search algorithm to optimize the spectrogram.
The step S1 specifically includes:
step S11: setting an energy judgment threshold of single-point spectrum sensing;
step S12: establishing a received signal energy set of all secondary users in a network at a certain moment;
step S13: comparing each element in the received signal energy set with an energy judgment threshold, if the signal energy is less than the judgment threshold, the secondary user judges that the target frequency spectrum is available, otherwise, the secondary user judges that the target frequency spectrum is unavailable;
step S14: forming a judgment result set according to the judgment results of all the elements;
step S15: and collecting the feature set formed by the judgment result set and the secondary user position information into a data fusion center.
The step S2 specifically includes:
step S21: constructing a constraint equation of a spectrum state boundary;
step S22: solving a constraint equation by a Lagrange multiplier method;
step S23: and determining a kernel function according to the solution result to construct a spectrogram.
The step S21 specifically includes:
step S211: constructing a spectrum state boundary formula in a sample space or a higher dimensional space:
wTφ(x)+b＝0
wherein: w ═ w1,w2,...,wN}TThe normal vector of the boundary of the spectrum state is obtained, N is a space dimension, b is a displacement parameter of the boundary of the spectrum state, phi (x) is a mapping function of a position information vector x on a higher-dimensional space, and if the position information vector x does not need to be mapped to the higher-dimensional space, phi (x) is x;
step S212: constructing a constraint equation of a normal vector w:
Figure BDA0001753249950000031
s.t.Sm＝wTφ(xm)+b+ξm,m＝1,2,...,M
wherein: i | · | purple wind2Is the square of the norm, ξmError between spectrum state and spectrum availability decision result divided for mth secondary user according to position information and spectrum state boundaryThe difference, C, is the pair error ξmM is the number of secondary users, SmSpectrum availability decision result, x, for mth secondary usermIs the location information vector of the mth secondary user.
The step S22 specifically includes:
step S221: will be provided with
Figure BDA0001753249950000032
Substituting into the step constraint equation to solve a Lagrange multiplier set α;
and step S222, obtaining a normal vector w and a displacement parameter b of the spectrum state boundary according to the Lagrange multiplier set α to obtain the spectrum state boundary.
The constraint equation after deformation in step S221 is:
Figure BDA0001753249950000033
Figure BDA0001753249950000034
wherein α is α1,α2,…αm,…,αM}。
The normal vector is:
Figure BDA0001753249950000035
the displacement parameter is determined by making the Lagrangian function
Figure 1
And taking the minimum value to obtain.
The kernel function is a radial basis function:
exp(-||xi-xj||2/2σ2)。
wherein: σ is the width parameter of the radial basis function.
The step S3 specifically includes:
step S31: setting an iteration step thresholdθfAnd a classification accuracy threshold thetaa；
Step S32: performing non-boundary user screening iteration;
step S33: feature set obtained after T-1 iteration
Figure BDA0001753249950000037
The corresponding secondary user is taken as a boundary user, where T is the number of iterations of step S32.
The step S32 specifically includes:
step S321: for lagrange multiplier subset after previous iteration
Figure BDA0001753249950000038
Feature set
Figure BDA0001753249950000041
And corresponding set of spectrum decision results
Figure BDA0001753249950000042
Set α lagrange multiplierst-1Sorting the smallest theta according to the absolute value of the theta from large to smallf×Mt-1Characteristic vectors and frequency spectrum judgment results of secondary users corresponding to the Lagrange multipliers are determined from the characteristic set Xt-1And a set of spectral decision results St-1Removing to obtain a removed feature set
Figure BDA0001753249950000043
Sum spectrum decision result set
Figure BDA0001753249950000044
Wherein: thetafFor iterative step threshold, Mt-1Representing the number of secondary users M needing to continuously monitor the target frequency band after the t-1 iterationtRepresenting the number of secondary users needing to continuously monitor the target frequency band after the t-th iteration;
step S322: using the feature set X obtained after the t-th iterationtAnd a set of spectral decision results StThrough the steps ofS2 calculating to obtain new Lagrange multiplier set
Figure BDA0001753249950000045
And spectral state boundaries wt Tφ(x)+bt＝0；
Step S323: for the obtained spectrum state boundary wt Tφ(x)+btThe historical data of the previous moment is used for testing when the value is 0, and the classification accuracy d is obtaineda；
Step S324: accuracy of classification daWith a threshold of classification accuracy thetaaComparison, if daLess than thetaaStep S33 is executed; otherwise, the iteration is continued.
Compared with the prior art, the invention has the following beneficial effects:
1) the method comprises the steps of establishing a frequency spectrum diagram in the large-scale cognitive radio network according to historical decision data based on a support vector machine model, and determining a frequency spectrum state boundary by utilizing the characteristic that a data sample in a binary classification problem of the support vector machine is linearly divisible in a sample space or a higher-dimensional space, so that the frequency spectrum state diagram capable of visually showing the availability of a target frequency band of the large-scale cognitive radio network is established.
2) Secondary users far away from the boundary of the spectrum state are screened out through a boundary user search algorithm, and the secondary users do not need to participate in detecting a target frequency band, so that the energy cost and the time cost for constructing the spectrogram are reduced.
Drawings
FIG. 1 is a schematic flow chart of the main steps of the present invention;
fig. 2 is a schematic diagram of a spectrogram construction strategy.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
Before proceeding with the detailed description of the present application, it is necessary to make the following explanation:
the main user is a user with authorized frequency band use authority
The secondary user is a user without the use permission of the authorized frequency spectrum, but has cognitive function and can opportunistically use the authorized frequency band on the premise of not influencing the communication of the primary user
The energy judgment threshold is an energy intensity threshold for judging whether the target frequency band is occupied, if the energy judgment threshold is smaller than the energy judgment threshold, the target frequency band is considered to be unoccupied, and the secondary user can use the frequency band.
A low-overhead real-time spectrogram constructing method based on history decision data learning, as shown in fig. 1 and 2, includes:
step S1: the method for loading the collected target spectrum availability historical decision data obtained by all sensing nodes in the large-scale cognitive radio network through a single-point spectrum sensing algorithm specifically comprises the following steps:
step S11: setting energy judgment threshold E of single-point spectrum sensingθ. Perceived signal energy intensity E for a certain secondary user mmIf there is Em＜EθThen the secondary user will determine that the target spectrum is available (not occupied by the primary user); otherwise, if there is Em≥EθThe secondary user will decide that the target spectrum is not available (already occupied by the primary user). Energy decision threshold EθCan be set according to different application environments and the requirement of system performance.
Step S12: establishing a received signal energy set of all secondary users in the network at a certain time:
E＝{E1,E2,...,Em,...,EM}，
wherein: emThe received signal energy of the secondary user M at this moment, E is the received energy set of all secondary users, and M is the number of secondary users.
Step S13: each element E in the energy set E of the received signalmAnd energy decision threshold EθComparing and making a corresponding spectrum availability decision, using SmAnd (4) showing. SmThe value of 0 or 1, 0 indicating that the spectrum is not mainly usedUser occupation, 1 indicates that the spectrum has been occupied by the primary user;
step S14: finally, judging the frequency spectrum availability of all secondary usersmEstablishing a frequency spectrum judgment result set S ═ S1,S2,...,Sm,...,SM}；
Step S15: forming a characteristic set X (X) by the position information (three-dimensional coordinates) of each secondary user1,x2,...,xm,...,xMIn which xmIs the three-dimensional coordinate vector of the secondary user m. And the feature set X and the judgment result set S are imported into a data fusion center for construction of a spectrogram.
Step S2: and analyzing the loaded historical decision data by using a least square support vector machine to determine a spectrum state boundary so as to construct a spectrogram, and based on a support vector machine model, considering that under the two-class problem, two types of samples can be linearly separable in a sample space or a certain higher-dimensional space, and the spectrum state boundary can be determined according to a relative distance maximum principle. The spectral state classification boundary determination process is as follows.
Step S21: according to the support vector machine model, constructing a constraint equation of the spectrum state boundary, which specifically comprises the following steps:
step S211: constructing a spectrum state boundary formula in a sample space or a higher dimensional space:
wTφ(x)+b＝0
wherein: w ═ w1,w2,...,wN}TThe normal vector of the boundary of the spectrum state is obtained, N is a space dimension, b is a displacement parameter of the boundary of the spectrum state, phi (x) is a mapping function of a position information vector x on a higher-dimensional space, and if the position information vector x does not need to be mapped to the higher-dimensional space, phi (x) is x;
step S212: and (3) constructing a constraint equation of a normal vector w, wherein the relative boundary distance gamma between each sample point in the sample space and the spectrum state classification boundary is 2/| | w | |. Making | | w | | non-woven phosphor according to the principle of maximum relative distance2The w value with the minimum value is the solution of the normal vector:
Figure BDA0001753249950000061
s.t.Sm＝wTφ(xm)+b+ξm,m＝1,2,...,M
wherein: i | · | purple wind2For operation of squaring norm ξmError between spectrum state divided by mth secondary user according to position information and spectrum state boundary and spectrum availability decision result, C is pair error ξmM is the number of secondary users, SmSpectrum availability decision result, x, for mth secondary usermIs the location information vector of the mth secondary user.
Step S22, solving a constraint equation by a Lagrange multiplier method, specifically, introducing a Lagrange multiplier subset α ═ α1,α2,...,αMSolving the constructed constraint equation by using a Lagrange multiplier method, wherein the Lagrange function converted from the constraint equation is
Figure BDA0001753249950000062
W, which causes the L (w, b, a) partial derivative to be 0, is the solution of the constraint equation, i.e.
Figure BDA0001753249950000063
The method specifically comprises the following steps:
step S221: will be provided with
Figure BDA0001753249950000064
Solving a Lagrange multiplier set α in the step constraint equation, and converting the constraint equation after substituting w into the step constraint equation
Figure BDA0001753249950000065
Figure BDA0001753249950000066
The constraint equation at this time is converted into linear constraint, and can be directly solved according to the constraint.
Step S222, obtaining a normal vector w and a displacement parameter b of the boundary of the spectrum state according to the Lagrange multiplier set α, and obtaining the boundary of the spectrum state, wherein the normal vector is as follows:
Figure BDA0001753249950000067
the displacement parameter is determined by making the Lagrangian function
Figure 2
Taking the minimum value, and the displacement parameter b can be obtained by making the Lagrangian function
Figure BDA0001753249950000069
And taking the minimum value to obtain. Thus, the spectral state boundaries are determined.
Step S23: determining a kernel function according to the solution result to construct a spectrogram, wherein the kernel function k (x) is determinedi,xj). The kernel function is defined as κ (x)i,xj)＝φ(xi)Tφ(xj) Wherein x isiAnd xjA location information vector for any two secondary users. In the support vector machine model, the kernel function k (x) can be directly selectedi,xj) Instead of the determination of the mapping function phi (x). In the spectral state boundary determination process described in this invention, the radial basis function exp (- | | x) is selectedi-xj||2/2σ2) As a kernel function, wherein: σ is a width parameter of the radial basis function, and is used for controlling the radial action range of the function.
Step S3: the invention determines support vectors in historical decision data through a boundary user search algorithm to optimize a spectrogram, and considers that in a large-scale cognitive radio network, a spectrum state boundary is mainly determined by historical data of secondary users (namely, boundary users) near the boundary, and secondary users (namely, non-boundary users) far away from the boundary have little influence on the determination of the spectrum state boundary. I.e. the importance of the border users is high, whereas the importance of the non-border users is low. Therefore, in the process of determining the boundary of the spectrum state, the non-boundary user can abandon the detection of the target frequency band without affecting the classification accuracy of the spectrogram, thereby reducing the energy overhead and the time overhead in the process of constructing the spectrogram. The boundary user search process in the invention is as follows:
step S31: setting an iteration step threshold thetafAnd a classification accuracy threshold thetaaIteration step thetaf(θfE (0,1)) is defined as theta in each secondary user participating in the iterationfMay be considered a non-boundary user. Threshold of classification accuracy thetaa(θaE (0,1)) is defined as the classification accuracy of the spectrogram after a certain iteration is less than thetaaAnd when the iteration process is ended, returning the result of the last iteration as a final result.
Step S32: performing non-boundary user screening iteration, specifically comprising:
step S321: for lagrange multiplier subset after previous iteration
Figure BDA0001753249950000071
Feature set
Figure BDA0001753249950000072
And corresponding set of spectrum decision results
Figure BDA0001753249950000073
Set α lagrange multiplierst-1Sorting the smallest theta according to the absolute value of the theta from large to smallf×Mt-1Characteristic vectors and frequency spectrum judgment results of secondary users corresponding to the Lagrange multipliers are determined from the characteristic set Xt-1And a set of spectral decision results St-1Removing to obtain a removed feature set
Figure BDA0001753249950000074
Sum spectrum decision result set
Figure BDA0001753249950000075
Wherein: thetafFor iterative step threshold, Mt-1Representing the number of secondary users M needing to continuously monitor the target frequency band after the t-1 iterationtIndicating the need to continue monitoring after the t-th iterationThe number of secondary users of the target frequency band;
step S322: using the feature set X obtained after the t-th iterationtAnd a set of spectral decision results StA new lagrange multiplier subset is calculated via step S2
Figure BDA0001753249950000076
And spectral state boundaries wt Tφ(x)+bt＝0；
Step S323: for the obtained spectrum state boundary wt Tφ(x)+btThe historical data of the previous moment is used for testing when the value is 0, and the classification accuracy d is obtainedaAccuracy of classification daIs defined as the accuracy with which the historical decision data at the previous time instant is classified by the spectral state boundaries.
Step S324: accuracy of classification daWith a threshold of classification accuracy thetaaComparison, if daLess than thetaaStep S33 is executed; otherwise, the iteration is continued.
Step S33: feature set obtained after T-1 iteration
Figure BDA0001753249950000083
The corresponding secondary user is taken as a boundary user, where T is the number of iterations of step S32.
And finally, selecting a simulation scene with the length of 30 kilometers and the width of 20 kilometers, and explaining the effectiveness of the boundary user search algorithm through simulation. In a simulation scene, the positions of main users with the transmitting power of 30mW are random, and the classification accuracy threshold theta isaSet to 0.95. The simulation results are shown in tables 1 and 2. Table 1 shows that when the number of secondary users is 500, the different iteration steps thetafThe following spectrogram construction performance; table 2 shows the iteration step θfAt 0.02, spectrogram construction performance is performed under different numbers of secondary users. Table 1 and table 2 show the effectiveness of the spectrogram construction method and the boundary user search algorithm proposed in this patent under different conditions, that is, the spectrogram has high classification accuracy and simultaneously reduces time overhead and energy overhead.
TABLE 1 iteration step θfFor spectrum graph structureInfluence of construction
Figure BDA0001753249950000081
Table 2 influence of secondary user number on spectrogram construction
Figure BDA0001753249950000082

Claims (8)

1. A low-overhead real-time spectrogram construction method based on historical decision data learning is characterized by comprising the following steps of:
step S1: loading collected target spectrum availability historical decision data obtained by all sensing nodes in the large-scale cognitive radio network through a single-point spectrum sensing algorithm,
step S2: the loaded historical decision data is analyzed using a least squares support vector machine to determine spectral state boundaries to construct a spectrogram,
step S3: determining a support vector in the historical decision data through a boundary user search algorithm to optimize a spectrogram;
the step S2 specifically includes:
step S21: a constraint equation for the boundary of the spectral state is constructed,
step S22: solving a constraint equation by a Lagrange multiplier method; ,
step S23: determining a kernel function according to the solving result to construct a spectrogram;
the step S21 specifically includes:
step S211: constructing a spectrum state boundary formula in a sample space or a higher dimensional space:
wTφ(x)+b＝0
wherein: w ═ w1,w2,...,wN}TIs a normal vector of the boundary of the spectrum state, N is a space dimension, b is a displacement parameter of the boundary of the spectrum state, phi (x) is a mapping function of the position information vector x on a higher-dimensional space, if the mapping is not needed to be mapped to the higher-dimensional space, phi (x) is x,
step S212: constructing a constraint equation of a normal vector w:
Figure FDA0002307052450000011
s.t.Sm＝wTφ(xm)+b+ξm,m＝1,2,...,M
wherein: i | · | purple wind2Is the square of the norm, ξmError between spectrum state divided by mth secondary user according to position information and spectrum state boundary and spectrum availability decision result, C is pair error ξmM is the number of secondary users, SmSpectrum availability decision result, x, for mth secondary usermIs the location information vector of the mth secondary user.
2. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 1, wherein the step S1 specifically comprises:
step S11: setting an energy judgment threshold of single-point spectrum sensing;
step S12: establishing a received signal energy set of all secondary users in a network at a certain moment;
step S13: comparing each element in the received signal energy set with an energy judgment threshold, if the signal energy is less than the judgment threshold, the secondary user judges that the target frequency spectrum is available, otherwise, the secondary user judges that the target frequency spectrum is unavailable;
step S14: forming a judgment result set according to the judgment results of all the elements;
step S15: and collecting the feature set formed by the judgment result set and the secondary user position information into a data fusion center.
3. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 1, wherein the step S22 specifically comprises:
step S221: will be provided with
Figure FDA0002307052450000021
Substituting into the step constraint equation to solve a Lagrange multiplier set α;
and step S222, obtaining a normal vector w and a displacement parameter b of the spectrum state boundary according to the Lagrange multiplier set α to obtain the spectrum state boundary.
4. The method as claimed in claim 3, wherein the constraint equation after deformation in step S221 is as follows:
Figure FDA0002307052450000022
Figure FDA0002307052450000023
wherein α is α1,α2,…αm,…,αM}。
5. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 3, wherein the normal vectors are as follows:
Figure FDA0002307052450000024
the displacement parameter is determined by making the Lagrangian function
Figure FDA0002307052450000025
And taking the minimum value to obtain.
6. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 1, wherein the kernel function is a radial basis function:
exp(-||xi-xj||2/2σ2)
wherein: σ is the width parameter of the radial basis function.
7. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 3, wherein the step S3 specifically comprises:
step S31: setting an iteration step threshold thetafAnd a classification accuracy threshold thetaa；
Step S32: performing non-boundary user screening iteration;
step S33: feature set obtained after T-1 iteration
Figure FDA0002307052450000031
The corresponding secondary user is taken as a boundary user, where T is the number of iterations of step S32.
8. The method for constructing a low-overhead real-time spectrogram based on historical decision data learning of claim 7, wherein the step S32 specifically comprises:
step S321: for lagrange multiplier subset after previous iteration
Figure FDA0002307052450000032
Feature set
Figure FDA0002307052450000033
And corresponding set of spectrum decision results
Figure FDA0002307052450000034
Set α lagrange multiplierst-1Sorting the smallest theta according to the absolute value of the theta from large to smallf×Mt-1Characteristic vectors and frequency spectrum judgment results of secondary users corresponding to the Lagrange multipliers are determined from the characteristic set Xt-1And a set of spectral decision results St-1Removing to obtain a removed feature set
Figure FDA0002307052450000035
Sum spectrum decision result set
Figure FDA0002307052450000036
Wherein: thetafFor iterative step threshold, Mt-1Representing the number of secondary users M needing to continuously monitor the target frequency band after the t-1 iterationtRepresenting the number of secondary users needing to continuously monitor the target frequency band after the t-th iteration;
step S322: using the feature set X obtained after the t-th iterationtAnd a set of spectral decision results StA new lagrange multiplier subset is calculated via step S2
Figure FDA0002307052450000037
And spectral state boundaries wt Tφ(x)+bt＝0；
Step S323: for the obtained spectrum state boundary wt Tφ(x)+btThe historical data of the previous moment is used for testing when the value is 0, and the classification accuracy d is obtaineda；
Step S324: accuracy of classification daWith a threshold of classification accuracy thetaaComparison, if daLess than thetaaStep S33 is executed; otherwise, the iteration is continued.
CN201810875938.6A 2018-08-03 2018-08-03 Low-overhead real-time spectrogram construction method based on historical decision data learning Active CN109167642B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810875938.6A CN109167642B (en) 2018-08-03 2018-08-03 Low-overhead real-time spectrogram construction method based on historical decision data learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810875938.6A CN109167642B (en) 2018-08-03 2018-08-03 Low-overhead real-time spectrogram construction method based on historical decision data learning

Publications (2)

Publication Number Publication Date
CN109167642A CN109167642A (en) 2019-01-08
CN109167642B true CN109167642B (en) 2020-04-28

Family

ID=64898841

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810875938.6A Active CN109167642B (en) 2018-08-03 2018-08-03 Low-overhead real-time spectrogram construction method based on historical decision data learning

Country Status (1)

Country Link
CN (1) CN109167642B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101867424A (en) * 2010-06-25 2010-10-20 西安交通大学 Cooperative frequency spectrum sensing method for cognitive radio network
CN109302250A (en) * 2018-09-13 2019-02-01 西安交通大学 Energy transmission full duplex relaying carries out relay selection and power distribution method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101867424A (en) * 2010-06-25 2010-10-20 西安交通大学 Cooperative frequency spectrum sensing method for cognitive radio network
CN109302250A (en) * 2018-09-13 2019-02-01 西安交通大学 Energy transmission full duplex relaying carries out relay selection and power distribution method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Historical Spectrum Sensing Data Miningfor Cognitive Radio Enabled Vehicular Ad-Hoc Networks;Xin-Lin Huang;《IEEE》;20160228;第13卷(第1期);1-12 *

Also Published As

Publication number Publication date
CN109167642A (en) 2019-01-08

Similar Documents

Publication Publication Date Title
He et al. Deep learning based energy efficiency optimization for distributed cooperative spectrum sensing
CN112512069B (en) Network intelligent optimization method and device based on channel beam pattern
Zhang et al. MIMO spectrum sensing for cognitive radio-based Internet of things
Ahmed et al. Deep learning-driven opportunistic spectrum access (OSA) framework for cognitive 5G and beyond 5G (B5G) networks
CN109525994B (en) High-energy-efficiency spectrum sensing method based on support vector machine
CN108337645B (en) Gridding radio signal monitoring system architecture and interference signal joint detection method
CN110224771B (en) Spectrum sensing method and device based on BP neural network and information geometry
CN107071788B (en) Spectrum sensing method and device in cognitive wireless network
Jaglan et al. Scalable and robust ANN based cooperative spectrum sensing for cognitive radio networks
Goyal et al. Deep learning application for sensing available spectrum for cognitive radio: An ECRNN approach
Zecchin et al. Robust Bayesian learning for reliable wireless AI: Framework and applications
CN110649982B (en) Double-threshold energy detection method based on secondary user node selection
CN109167642B (en) Low-overhead real-time spectrogram construction method based on historical decision data learning
CN104158604B (en) A kind of distributed collaborative frequency spectrum sensing method based on average common recognition
CN111934797B (en) Collaborative spectrum sensing method based on covariance eigenvalue and mean shift clustering
De Souza et al. Compressive learning in communication systems: a neural network receiver for detecting compressed signals in OFDM systems
CN110351884B (en) Spectrum opportunity access method based on double-layer multi-arm tiger machine statistical model
CN108490877B (en) Diagnosing faults of numerical control machine method based on symmetrical dot pattern and space density clustering
Syed et al. Deep learning approaches for spectrum sensing in cognitive radio networks
Li et al. Accurate Anomaly Detection with Energy-Efficiency in IoT-Edge-Cloud Collaborative Networks
CN113747441A (en) Mobile network flow abnormity detection method and system based on feature dimension reduction
Liu et al. Adversarial training for low-complexity convolutional neural networks using in spectrum sensing
Yaseen et al. Graph signal processing-based network health estimation for next generation wireless systems
Liu et al. Topology sensing of wireless networks based on Hawkes process
Bharathy et al. Gini index test metric based collaborative sensing of spectrum for cognitive networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant